@saehrimnir/druidjs / ParametersPaCMAP
Interface: ParametersPaCMAP
Defined in: dimred/index.js:222
Properties
| Property | Type | Description | Defined in |
|---|---|---|---|
apply_pca? | boolean | Reduce inputs wider than 100 dimensions to 100 via PCA before the search and the initialization. | dimred/index.js:234 |
d? | number | the dimensionality of the projection. | dimred/index.js:226 |
FP_ratio? | number | Further pairs per point, as a multiple of n_neighbors. | dimred/index.js:225 |
knn? | | KNN<number[] | Float64Array<ArrayBufferLike>, any> | null | Index used to find the neighbors. If null, an exact blocked search runs instead, which is faster than a tree at the n_neighbors + 50 candidates the density rescaling needs. Pass an approximate index such as HNSW, Annoy, or NNDescent to avoid the O(N^2) search on large datasets. Default is null | dimred/index.js:230 |
lr? | number | learning rate of the Adam optimizer. | dimred/index.js:228 |
metric? | Metric | the metric which defines the distance between two points. | dimred/index.js:227 |
MN_ratio? | number | Mid-near pairs per point, as a fraction of n_neighbors. | dimred/index.js:224 |
n_neighbors? | number | Number of nearest neighbors forming the attractive pairs. | dimred/index.js:223 |
num_iters? | number[] | Iterations in each of the three phases. | dimred/index.js:229 |
seed? | number | the seed for the random number generator. | dimred/index.js:236 |